Depression and the Link Between Radiographic OA, Knee Pain
In our study, a high proportion of participants had radiographic findings but did not report knee pain (21.3%). This is in accordance with previous data. Pain is the most frequent complaint in OA and frequently the first reason for consulting a physician. It is an unspecific symptom and its expression may be associated with other conditions than OA, making pain assessment a relevant but difficult issue. We found that participants with higher pain scores had higher odds of having depressive symptoms. Additionally, a higher association of knee pain with radiographic knee OA was found in participants without depressive symptoms. The probability of a participant having radiographic KL ≥ 2 rose substantially with the number of positive answers to knee pain questions (increased pain scores). This is more obvious in score 2 and 3 where the higher likelihood ratios were found. These results reinforce our idea that using a score was better than using each separate question and they were in accordance with previous studies showing a lower ability of individual questions on knee pain for the diagnosis of OA.
Our results were consistent in females and males, but positive answers in males contributed to higher likelihood ratios than in females. This may be explained by sex differences in pain perception, evaluation, and reporting, with females reporting pain more frequently than males, which may make these symptoms more unspecific in females.
Since, in general, our data showed similar results by sex, and in order to improve the statistical power, we decided to analyze the role of depressive symptoms not stratifying by sex. Among participants with BDI ≤ 14 the likelihood ratio to identify patients with radiographic knee OA increased with increased pain scores reaching 7.34 when participants responded positively to all pain questions (score 3). In the presence of depressive symptoms (BDI > 14) our score became unable to identify participants with radiographic knee OA since the likelihood ratios ranged from 0.77 for those with one positive answer (score 0) to 1.82 for those with all positive answers (score 3). These results are in accordance with other studies that show that depressive symptoms can change pain perception and contribute to pain over-expression. Thus, in the presence of depressive symptoms, pain becomes more unspecific so the ability of pain complains to allow the identification of patients with radiographic OA is lower.
It is important to highlight that our data was from a population-based study and participants had low depressive symptomatology [median of 7 (25th–75th percentile: 3–12)]. It can be argued that in populations with more prevalent depressive symptoms, such as older populations or those with higher co-morbidities, knee pain can even have a lower discrimination ability to identify participants with OA.
On the other hand, among participants with depressive symptoms, pain questions could be useful to identify those without disease, since negative answers (reporting no pain) allowed the best specificity (57%) and a positive likelihood ratio of 0.50.
Considering all participants, the score −1, corresponding to participants that reported no pain, was unable to efficiently discriminate participants according to radiographic OA findings (it presented a high sensitivity, but a very low specificity and the lowest likelihood ratio). So, reporting "no knee pain" was not enough to accurately identify participants without radiographic OA. However, stratified analysis by depressive symptoms, showed that among participants with depressive symptoms the specificity of score −1 rose to almost 60%, showing that in this group of participants reporting "no knee pain" may be a better marker to identify participants without radiographic OA.
Pain related to OA is associated with an increased risk of worse psychological outcomes, including anxiety, depression and helplessness. Since our data is based on a cross-sectional study, it is impossible to understand if depressive symptoms are responsible for OA pain manifestations or whether they are caused by OA pain. However, this was not the purpose of our study. We tried to understand if the presence of depressive symptoms, regardless of their causes, may influence knee pain expression and its ability to identify patients with relevant radiographic OA findings.
We tried to minimize the problem of pain assessment using questions with different time frames. Nevertheless, the ability of knee pain to identify participants with radiographic OA would probably improve with a higher number of questions; however, we tried to use a small number of easy questions, which could be easily used in a clinical setting or in population based studies.
One limitation of our study was the inability to explore other common features of OA such as stiffness, loss of joint mobility, swelling, tenderness, joint deformity and muscle weakness. However, pain is thought to be one of the aspects that is more correlated with radiographic symptomatic changes and is also highly associated to the other OA signs and symptoms, namely disability. So we could expect a similar effect of depressive symptoms on those signs and symptoms.
In our study radiographic OA was defined as Kellgren Lawrence score of 2–4, and some reliability and validity limitations of this grading scale can be discuss. Our radiographs were scored only by one experienced reader that was unaware of the participants' clinical data. We used a cut off point ≥ 2 to define radiographic OA which is also a controversial aspect. Since differences would be expected in the relation between pain and radiographic changes according to disease severity, we performed the same analyses using KL ≥ 3 as cut off point and we found a slight improvement in the ability of knee pain to identify participants with radiographic OA, but those differences disappeared among those with depressive symptoms (data not shown).
Although our study was developed from a population-based study, we lost almost 33% of the participants for our research, which could determine a selection bias. Nevertheless, non-participants were quite similar to the studied population although younger. Thus, despite some limitations, our results support the idea that psychological status deserves greater clinical and research attention in OA, since the presence of depressive symptoms impairs the ability of pain complaints to identify potential OA patients or to correctly manage the disease in those already diagnosed. The evaluation and treatment of depressive symptoms should be an integrative part of OA patient management, in order to improve it and to ensure that knee pain expression can be correctly understood.